example

Serverless Plugin Canary Deployments Example

Let's see the plugin in action 😄 In this service example we'll create several resources:

• A Lambda function with an HTTP event where we'll test the gradual deploy
• A before and an after traffic shifting hooks
• An alarm that sets off if the Lambda function exits with error status

The function's deployment is configured so that it shifts a 10% of the traffic to the new version every minute, as long as the alarm status is OK.

Usage

First, we need to set up the service and deploy it in our AWS account:

$ npm i
$ sls deploy -s dev

When you call your newly created endpoint, you should see the following message:

$ curl https://yourendpoint.com/dev/hello
{"message":"First version"}

To check how traffic is shifted gradually, modify handler.js and deploy your service again. You'll see that the output varies accross endpoint calls.

$ sls deploy -s dev
$ curl https://yourendpoint.com/dev/hello
{"message":"First version"}
$ curl https://yourendpoint.com/dev/hello
{"message":"Second version"}

If we wanted to check how our deployment rolls back to the previous version, we need to set off the alarm. Calling the endpoint with error as a query parameter will cause the function with error status.

$ curl https://yourendpoint.com/dev/hello?error=true
{"message": "Internal server error"}

As soon as the alarm turns to ALARM state, all the traffic will be shifted to the previous version. You can check progress of the deployment in the CodeDeploy console.